The technology inside the Google self-driving cars

You’ve probably heard that Google has these cool cars that can drive themselves. Above is a video of one of them driving down Freeway 280 that I shot almost a year ago (I had no idea I was driving next to a car that was capable of driving itself. I should have known something secret was going on because the driver braked hard to get away from my camera). Yes, there’s a driver behind the wheel, but the car is capable of driving itself.

But Google is very closed-lipped about what these cars are actually doing. I’ve asked several times, to several different people, to get a ride in one, and to be able to interview the team about what they are doing. I keep getting turned away.

I’ve learned a few things, though, in the last few weeks. For one, Mike Montemerlo is the brains behind many of the algorithms that run the car. I interviewed him back in 2007. When I talk with folks familiar with the Google program they all say he knows more about the code running the car than anyone else, so this interview is important to go back to and watch. You can see a real visionary at the beginning of his journey to invent the future.

Recently I went back to Stanford to visit the Center for Automotive Research at Stanford (which, funnily enough, spells “CARS”). This is the coolest place for a geek. You want to inspire your kids to learn computers? Show them what the cars of the future will need. Hint: computer scientists.

In Part One of the tour, you get to see the latest brilliant kids who are pushing cars to do things even race cars can’t do. But in Part Two, you get to see the technology that is inside the Google self-driving cars (the Google team came from Stanford, and the approaches are very similar and the two teams keep pretty close ties). In Part Three of the tour, coming next week, you’ll get a look at the solar car team.

So, what did I learn from Mike Sokolsky, who is a research engineer in the artificial intelligence lab at Stanford? Well, they are studying ways to process a massive amount of data (there’s a tiny datacenter inside the back of the car) and make real-time decisions. Was that a stoplight? Was that a cat? A ball? A child?

One thing he ends on, is the challenges ahead for getting this technology into production cars. He thinks it’ll be more than 10 years. Why?

Well, most humans don’t like the idea of a car that drives them around. They like control, and this car would take control away from them.

There’s a deep set of insurance issues, too. What happens if one of these cars causes a wreck? Who is responsible?

Of course there’s also the cost. The Stanford car has about $300,000 worth of gear inside of it. It’ll take 15 to 20 years to reduce the cost of the gear to $3,000 or less (I remember about 20 years ago Steve Wozniak proudly showed me his dye sublimation color printer. It cost him $40,000 in 1990 dollars. Today a better printer costs about $50).

Funny story, the LIDAR unit on top of the car? It was invented by David Hall. His Silicon Valley-based company, Velodyne, used to make subwoofers. I sold those in my consumer electronics store in Silicon Valley. A few weeks back his wife called me and told me a bit about him. I love that people don’t understand why I put my cell phone number on my blog — this is exactly why. It’s +1-425-205-1921 by the way. In the video we did at Stanford you can see that the LIDAR shows how it maps the world in 3D 10 times a second.

Can these technologies do even better than humans can? Absolutely. Check out this video of a sliding move into a parking spot by the car we visited. I can’t do that. Can a race car driver? Yes. But can a race car driver have sensors that know exactly how far to push adhesion to the road? Not always. The Stanford engineers I’ve talked to say they are already able to do things that would be difficult for a race-car driver to do over and over again.

This technology is the most mind-blowing stuff out there now. It’s like seeing the future. Enjoy!

More details are on Building43. Thanks to Rackspace for letting me chase around world-changing technology like this. Do you have world-changing technology? Let us know at scobleizer@gmail.com.

Here’s a report from ABC News, which gives you a look inside the Google car:

About Robert Scoble

As Startup Liaison for Rackspace, the Open Cloud Computing Company, I travel the world with Rocky Barbanica looking for what's happening on the bleeding edge of technology and report that here.

9 thoughts on “The technology inside the Google self-driving cars

  1. Seems like a generation shift will need to take place before we see self-driving cars, 10 – 20 years seems about right. Besides, if I spit up 300K for a car it better do my laundry, file my taxes and take my dog for walk…before it even leaves the driveway….

  2. Correction: “Show them what the cars of the future will need. Hint: computer scientists.”My answer: Hint: Robotics scientists. You are going to need swarm robotics algorithms to make sure all cars on the road are able to drive/coordinate with each other while still obeying traffic laws. Computing is only part of the equation. Search for youtube videos on ‘swarm robotics’ and you will see what it can do. That’s what is needed for any gruop of vehicles that are going to drive themselves while taking care not to bump into each other. Too bad chaffeurs are going to look for some other job sometime soon. I even made a blog post about it on my blog yesterday but I am not sure if links are allowed here so if you want to read about it you can visit my blog and look for a post that reads ‘the connected brain’.

  3. Great but worried that Google are straying too far away from core products. With this and Larry and Sergy going to Mars I think they are going away from offering a great products to people that know what the brand does.

    Chatting with Chris Brogan there is discontent with Chrome, so with discontent with Buzz, Wave and Chrome that is alot of discontent. Sticking to attracting good people as you mentioned with Lars, and great products (maybe these are) would work out best.

    1. I disagree with you 100%. Why? Because by doing this kind of research Google is learning a lot about a whole range of fields that they would never learn about. For instance, how to handle such a large amount of real-time data. This will also help them build better maps and better turn-by-turn navigation.

      R&D is very important and it’s important that Google NOT be stuck in doing what worked for them yesterday. That’s why this project is so interesting to me. I want to see what spins off of it.

      Remember, the Web was invented at CERN. An atom-smasher. That wasn’t their core business, but it spun out and changed the world.

      If you hold organizations to only stick with what got them there, you’ll be severely limiting their potential.

      1. Or look at Bell Labs. There were lots of separate discoveries unrelated to their “core” business, but many of those fed back into improvements in AT&T’s product.

        Thinking about this post in contrast to your Google/Instagram post…I think Instagram is more an iterative product, while a self driving car is big league innovation. I don’t mean that as a slam on Instagram, but I can’t imagine a 2 person team bringing us this much closer to a self driving car.

        Maybe one day 3 people shops will put a man on the moon. The price for world changing innovation is plummeting. It’s not cheap yet, and until then it will take the larger teams available from established companies to create bedrock-type innovations.

        1. Actually, from what I hear, the Google self-driving car team was developed at Stanford and CMU, with very small focused teams. All the code at Stanford, for instance, was checked in by one guy. Why does that work? Because he knew how the entire thing worked and was more efficient than having a larger decentralized team. Eventually the teams take over, though. Remember the Apple II? Designed by one guy. Doing the Mac, though, took teams.

    1. Me too. I’m already using some assisted technologies in the latest Toyota Prius. But it’s not enough. I want a car that totally drives itself so I can blog instead of driving around town. :-)

Comments are closed.